Gene expression profiling indicates that at least 10,000-15,000 genes are actively expressed in human breast cancer. However, review of literature reveals that the number of genes actively researched in breast cancer is well below 1000. Therefore, a large part of the human breast cancer transcriptome still remains uncharacterized. The current paradigm of therapeutic target discovery starts with experiments in cancer models and promising candidates are taken to the clinic for testing. The advent of high-throughout molecular analytic techniques have enabled a paradigm shift in the discovery process by allowing the discovery and validation of new markers to be made by analyzing human cancer data. New strategies are required for therapeutic target discovery in cancer that address the complex genetic landscape of the disease. First, it is difficult to reproduce in model systems the genetic complexity of human cancers. Although large-scale sequencing studies have the potential to uncover mutated genes that could be diagnostic or therapeutic targets, undertaking such efforts by individual groups is cost prohibitive. Furthermore, due to the large heterogeneity in the mutated genes involved in a particular behavior of a tumor, large sample sizes will be required to reach statistically meaningful conclusions. Studies have shown that individuals that are heterozygous for a single nucleotide polymorphism (SNP) in a cancer gene show significant allelic variation in the expression of the corresponding mRNA. Therefore, genotypes can be connected to disease susceptibility based on changes in gene expression as opposed to differences in SNPs in the gene itself, which are considerably more difficult to detect. Most genes within a population of normal or cancer samples exhibit a broad range of expression values that reflect differences in the genetic regulation or potential allelic variation among individuals. Such gene transcript distributions are typically asymmetric and left-skewed with a few individuals having expression levels that are considerably greater than the median. Interestingly, two of the most important therapeutic targets and phenotype determinants of human breast cancers, the estrogen receptor a (ER-a) and the epidermal growth factor receptor 2 (HER-2), exhibit markedly bimodal distributions of expression within populations of primary breast tumors. Motivated by this observation, we hypothesize that genes exhibiting bimodal expression distributions within primary breast tumors might be important as markers. We propose to perform a genome- wide search to identify such markers and to characterize and prioritize them for further development in future studies. Breast cancer is a leading cause of mortality and suffering in women despite advances in early detection techniques. Identification and validation of new targets for therapeutic intervention is an ongoing effort requiring a lengthy process that typically starts with experiments in cancer models to identify the most promising candidates, which are eventually tested in the clinic. Although cancer is ultimately caused by several accumulated mutations in key genes, identifying such cancer causing mutations is very challenging due to the large variability between individual tumors. We propose to evaluate a new method that takes advantage of this variation in expression levels of genes in breast cancers to identify potentially important therapeutic targets. Two of the most important therapeutic targets in breast cancer, estrogen receptor (ER-a) and epidermal growth factor receptor 2 (HER-2) exhibit the same kind of variability among individual tumors offering a justification for the proposed strategy. [unreadable] [unreadable] [unreadable]